0
\$\begingroup\$

I am using the three of the RT9624A half bridge driver as part of a 3-phase motor control circuit. Everything was working well in testing until I got to about 80% duty cycle and one of the drivers started getting hot. Probing it found that the low-side output had gotten internally shorted to VDD.

I looked back at the datasheet of the RT9624A and noticed that all the example schematics include a resistor in series with the VCC, apparently limiting current into the IC. Why is this necessary and is it always necessary? Could this have something to do with my failure? The datasheet does not seem to explain. I also notice that the datasheet warns that too small a bootstrap capacitance can lead to overcharging and damage the chip. They give a minimum of 0.1uF, and recommend 1.0uF. I have 0.47uF, which seems fine to me, but could I need a higher value?

This is my circuit, with GHA and GLA directly wired to 51nC total gate charge N-type mosfets. At 12V, this makes 4.25nF, and with a PWM frequency of 20kHz, switching losses should be less than 100mW. The part handles up to 800mW, so that shouldn't be the issue. enter image description here

Edit: Although I haven't confirmed anything, I have several action items to prevent similar issues in the future.

  1. Adding gate resistors, and also fattening and shortening the gate traces to limit potential harm from inductance loops.

  2. Similarly reducing inductance of the SW/PHASE trace, and adding a schottky diode on the SW/PHASE node to prevent damage from negative voltage spikes. Currently my mosfet's body diode will be limiting those spikes to at worst -1.1V, but the RT9624A can only handle voltages below -0.3 for at most 100ns, so extra protection will help. If anything, based on my probing my circuit, I think it was low voltage produced by motor back EMF that might have caused the damage.

This different datasheet says "Pulling HO more than –0.3 V below HS can activate parasitic transistors resulting in excessive current flow from the HB supply, possibly resulting in damage to the IC. The same relationship is true with LO and VSS. If necessary, a Schottky diode can be placed externally between HO and HS or LO and GND to protect the IC from this type of transient."

The IR2301 specifically notes it is "Tolerant to negative transient voltage dV/dt immune" and says the VS/SW/PHASE pin can go as low as the bootstrapped voltage -25V, which in my case would be something like -8V. When dealing with a BLDC/3-phase controller, this is probably an additional feature that could be very useful. (and saved me this problem).

  1. This guide gives the motivation behind using a series resistor to not only bypass noise but also decouple the chip from the rest of the circuit. Accordingly, I will also be adding in the 2.2Ohm example resistance from the datasheet in between the battery power and VCC. I do find it interesting, though, that in looking at more than 10 other half bridge datasheets, not a single other included a series resistor on VCC.
\$\endgroup\$
2
\$\begingroup\$

That resistor in series isn't just a resistor in series. It is combined with that capacitor to form a low-pass filter.

Notice that the resistor has a very low value in their example, just about 2 ohms. When the gate driver turns on the MOSFETs, it is going to suddenly draw a lot of current as it attempts to charge the gates as fast as possible. The resistor forces it to take it from that capacitor and isolates the VCC line from resonating/ringing. You want a larger capacitor so that your VCC takes longer to dip down. Since the VCC's sudden draw of current only happens for a short time, this will reduce the amount that VCC drops. Remember that with an RC filter, that capacitor will take time to recharge after VCC briefly discharges and so if you make your capacitor too large, it will take a longer time to recharge.

Now, what happens if you don't have that resistor? Your driver will first draw from the capacitor, but the sudden inrush of current makes its way backwards along the power rail in the form of a voltage dip. When that dip hits the source (your regulator or another capacitor), it will likely reflect and become higher as it returns back to the driver's VCC pin. This oscillation could persist for some time and may lead to destructive voltages on the part, depending on the speed (frequency) of the draw on VCC and the geometry of the connection to your power rail (most every piece of copper is a a non-negligible inductor at 10's of ns gate speeds). Adding the resistor in series damps this oscillation (with a corner frequency of \$1/(2\pi RC))\$. In short, that resistor should be considered always necessary, though the value may vary and could depend on the capacitor value you choose.

Another issue you might face is with the length of the trace to the gates from your driver. You must do anything you can to minimize the inductive loop between the gate signals, the GND terminal, and the SW terminal. Keep the copper carrying the signals short, fat, and as close to the other signals as possible. If you don't you'll get the same sort of inductive ringing that you could see on your VCC line and that too could destroy your part.

I would recommend probing at two places: Your Vcc and your two gate signals. You might immediately notice the problem. Based on your failure mode I would suspect ringing on one of your gate lines is doing your driver in.

I also would recommend simulating this circuit with parasitics as well if you have an accurate spice model of the RT9624A that includes VCC inrush. Represent the connection from your power rail to VCC with a small inductor (like 1nH or some other calculated value). Put small inductors on your gate lines (like 1nH or some value you calculate). Just see what happens.

\$\endgroup\$
  • \$\begingroup\$ I just probed those locations with my scope and didn't see anything suspicious. (both locations agreed with theory, dropping in voltage when more current was flowing, without any spikes). However, I was just using a 7.2V nominal battery, whereas the failure occurred while using a 11.1V nominal battery. +V_MOTOR is in fact direct from the battery, and so shouldn't have the kinds of problems output from a regular would have. I'll try to probe again for the problem while using a higher voltage battery. I have yet to examine the test cases you mention in spice. \$\endgroup\$ – Akh Jul 11 '18 at 4:33
  • \$\begingroup\$ Also, I am using this part at a drastically lower frequency, 20kHz, than the max of say 1MHz. In this case is the series resistor still strictly necessary? \$\endgroup\$ – Akh Jul 11 '18 at 4:38
  • 1
    \$\begingroup\$ It is. It isn't the switching frequency, it's how fast the gates turn on. Though the lower switching frequency would allow you to use a larger capacitor without too much consequence. Also, a battery will also present an impedance at high frequencies (which is why batteries are coupled with capacitors) and you will definitely see reflections. In fact, while batteries can usually source more DC current than a regulator they are often quite bad at fast current changes unless used with sufficient capacitance to present a lower impedance at higher frequencies.. \$\endgroup\$ – Los Frijoles Jul 11 '18 at 4:50
  • 1
    \$\begingroup\$ Also, double check that you are very far away from your Vgs rating on your gate signals. That is a hard limit. Something that could happen is that you perforate the gate of your mosfet and your driver ends up driving your motor. \$\endgroup\$ – Los Frijoles Jul 11 '18 at 4:54
  • \$\begingroup\$ Yeah, if you haven't already, check if the low-side FET is still good on the bad board. \$\endgroup\$ – mkeith Jul 11 '18 at 5:13
1
\$\begingroup\$

Take a look at figure 2 in the datasheet. Notice there is a driver for the low side powered by VCC. What you are saying is that the internal driver output is shorted to VCC. Probably what happened is that LGATE (aka DRVL aka GLA) got driven to a high voltage and caused the low-side driver output to short. Probe LGATE on a working unit and slowly raise the duty cycle to see if you have some high voltage spikes on LGATE.

Why would this happen? When the high side turns on, the phase (MOTA) voltage rises very rapidly (high dV/dt). The parasitic capacitance from drain to gate of the low-side FET can cause injection of a lot of current to the LGATE (aka DRVL aka GLA) pin. Adding a resistor in series with the low-side gate can help reduce any current injected that way. Probing across that resistor can allow you to watch for that current injection event.

\$\endgroup\$
  • \$\begingroup\$ Okay, looking at the typical application circuit, they end up with at least 2.2Ohm in series with each path from a voltage source to the mosfet gate. For the high gate, it is close to the mosfet gate. For the low gate, it is on VCC. The high gate also has a 1Ohm resistor by the bootstrap capacitor, but they leave this one off more often in the examples than the others. \$\endgroup\$ – Akh Jul 11 '18 at 5:30
  • 1
    \$\begingroup\$ Look for some datasheets from other vendors (like fairchild) for their half-bridge drivers. They explain a lot more stuff. This richtek datasheet is crap. Richtek parts are fine, but this particular datsheet sucks. \$\endgroup\$ – mkeith Jul 11 '18 at 5:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.